Abstract
Aiming at the high cost, poor anti-interference and the limitations of traditional stereo matching algorithms in small object detection, an algorithm integrating aanet neural network in the matching process is proposed. Zhang Zhengyou calibration method and Bouguet algorithm are used for camera calibration and binocular epipolar correction respectively, so as to obtain highly parallel binocular images; Aanet neural network is used for stereo matching to obtain accurate parallax information; The three-dimensional coordinates of the target point are calculated by combining parallax information and camera parameters. Experimental results show that the parallax image obtained by this algorithm has high frame rate and matching accuracy, and the error of three-dimensional coordinates is very small.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call